nGrams

Uncharted can be thought of as a case study for a piece of software that demonstrates two emerging intellectual trends: big data and digital humanities. These are explored in the book though the creation of the Ngram Viewer interface for examining the scanned Google Books collection.

The rich availability of ancient Greek texts in the Thesaurus Linguae Graecae (TLG) has opened up new types of research for Biblical and Patristic scholars. A very helpful feature on the TLG's website is the option to trace quotations with the help of an n-grams module (Intertex-tual Phrase Matching). However, it is virtually unknown how well this module performs and what scholars might expect from the results it produces. The core of this article, therefore, is devoted to a critical examination of the algorithm and of its results. The gospel according to John has been compared with the Paedagogus of Clement of Alexandria, with the gospel according to Matthew, and with the complete works of Plutarch. As it turns out, the algorithm performs well in cases of longer quotations with no or very few interpolations. Short quotations, however, are missed while interpolated or adapted quotations are poorly handled. It is suggested that the algorithm might perform better if the team of the TLG were to revisit its decision to ignore stopwords and if the algorithm were to allow for foreign words in its n-grams. Finally, it is advised that more transparency in the algorithm's mechanisms and a possibility for manually tuning its parameters might improve its applicability.

Development of text to speech and speech recognition technology in Tamil shall make interaction with the computing devices directly in Tamil eminently possible for the common
man, even if he is not technically savvy. This will make equal access to knowledge for everybody. While viable text to speech technology in Tamil is already available, there is hardly any research or development work going on in automated speech recognition (ASR) in Tamil. The conventional speech recognition paradigms that have been used for the European languages is ill suited for Tamil, since Tamil is a morphologically very rich language. The only speech recognition technology that currently is reasonably successful that is also used in commercial ASR systems makes use of a huge vocabulary and is essentially a transcription system that finds the sequence of words from the fixed vocabulary that best matches the input speech signal. This model cannot be directly used for most of the Dravidian languages such as Tamil, since these languages possess an exponentially growing lexicon. While the current commercial English ASRs employ a lexicon of about two lakh words, our own work has
identified over sixteen lakh unique words in Tamil and the size is still growing. This article proposes an alternate approach to speech recognition that must be explored to develop workable recognition technologies for Tamil.

Television drama, despite its enormous popularity across the globe, has rarely received attention from the linguistics field. The dearth of research into television drama dialogue is further exposed by the thriving contributions from various other fields such as philosophy, psychology, cultural studies and media studies. This paper seeks to promote research interest in this unique mediated text by selecting the renowned medical dramedy House M.D. as research subject and comparing its 927,922-word House M.D. pure dialogue corpus (HMDC) to both the 450-million-word Corpus of Contemporary American English (COCA) and its 95-million-word spoken subcorpus (COCA Spoken) using an adaptation of Bednerak's (2011) ranked frequency list method. Using WordSmith Tools in the calculation of n-gram (n = 1, 2, 3) at the words/clusters level, the findings indicate that HMDC is more interpersonal than COCA and has a closer resemblance to COCA Spoken than to COCA. HMDC also contains 3.4 times more negativity than COCA Spoken and 2.8 times more than COCA. As such, viewers are presented with English far more interpersonal, as well as involving significantly more disagreement than one will encounter in the real world. This study not only shows similarities and differences between House M.D. and contemporary American English, but also provides a preview of the huge potential in television drama-related research.

A Tweeter is the Social Media Network to demonstrate the different kind of language which having an independent nature of classifiers, presenting an result on the several text classification. A classification problems text general classification and topic detection in several language forms like Greek, English, Dautsch and Chinese. Then the study on key factors in the CAN (i.e. Chain Augmented Naive) model that can influence the classification performance of the global context and local context. Two novel smoothing techniques variation of Jelinek-Mercer and linear inter polation technique which perform existing methods. Natural languages are full of collocations, recurrent combinations of words that occur more often than expected by chance and that correspond to arbitrary word usages. Recent work in lexicography indicates that collocations are in English apparently they are common in all types of writing, including both technical and nontechnical generations. These kind of document describes the properties and some applications of the Microsoft Web Ngram corpus. The corpus can have the characteristics, contrast to static data distribution of previous corpus releases, this N-gram corpus is made publicly available as an XML Web Service so that it can be updated as deemed necessary by the user community to include new words and phrases constantly being added to the Web.

O Google Ngram Viewer é uma ferramenta online que permite a visualização, em gráficos, da frequência de aparição de grupos de até cinco palavras em livros do acervo Google Books. Este trabalho utiliza a ferramenta para investigar a presença bibliográfica de termos significativos da história do design, abrangendo seis áreas temáticas: personalidades do design, denominações do design gráfico, tipografia, tecnologias de impressão, design social e emocional, e teoria do design. A partir da visualização de ocorrências em determinados períodos, e sua contextualização, este trabalho demonstra possibilidades de uso da ferramenta quando aplicada no estudo da história do design. Assim, este trabalho procurou, através da visualização de ocorrências de termos relevantes para diferentes aspectos da atividade, identificar situações que levantassem questões interessantes. O objetivo da análise dos gráficos resultantes das buscas não foi encontrar respostas absolutas, mas sim perguntas e indicações para possíveis explicações e desenvolvimentos. Acreditamos assim que o Ngram Viewer se apresenta como uma ferramenta relevante para o desenvolvimento de pesquisas na área do design. Palavras-chave: visualização de dados; Ngram viewer; Google Books; história do design; gráficos.
Abstract
Google Ngram Viewer is an online graphing tool that charts the count of occurrences of groups of up to five words in the Google Books database. This work employs the tool to investigate the literary presence of significant terms in design history, including six topics: design personages, graphic design denominations, typography, printing technologies, social and emotional design, and design theory. From the visualization and contextualization of occurrences, this work describes possibilities of applying the tool in the study of design history. Therefore, this work intends to identify ― through visualization of the occurrences of terms, relevant to different aspects of our field ― situations that could raise interesting questions. The goal of the analysis of graphs, results of searches, is not to find absolute answers, but indeed questions and indications leading to possible explanations and developments. We believe, therefore, that Ngram Viewer presents itself as a relevant tool for research development in the design field.

This article is an experiment in applying some techniques from text analysis and network analysis to book history, specifically to the study of the title catalogs of four libraries in Germany in the 17th and 18th centuries. My goal is to make empirical discoveries about the collections that would be difficult or impossible to do through purely reading and study of the 1,296 titles contained in the four collections. First, I briefly discuss the domain of work – in time and space, and the model of library collections I employ. Second, I account for the methods I employ and introduce the theory of distant reading that informs the study. Third, I present findings, in several areas – general findings, findings about law libraries in Marburg, and findings about the city library of Frankfurt. I contend that, despite the limitations of the methods, some real contributions to our knowledge about these collections and perhaps others like them can be made with these methods.

This work extends the set of works which deal with the popular problem of sentiment analysis in Twitter. It investigates the most popular document ("tweet") representation methods which feed sentiment evaluation mechanisms. In particular, we study the bag-of-words, n-grams and n-gram graphs approaches and for each of them we evaluate the performance of a lexicon-based and 7 learning-based classification algorithms (namely SVM, Naïve Bayesian Networks, Logistic Regression, Multilayer Perceptrons, Best-First Trees, Functional Trees and C4.5) as well as their combinations, using a set of 4451 manually annotated tweets. The results demonstrate the superiority of learning-based methods and in particular of n-gram graphs approaches for predicting the sentiment of tweets. They also show that the combinatory approach has impressive effects on n-grams, raising the confidence up to 83.15% on the 5-Grams, using majority vote and a balanced dataset (equal number of positive, negative and neutral tweets for training). In the n-gram graph cases the improvement was small to none, reaching 94.52% on the 4-gram graphs, using Orthodromic distance and a threshold of 0.001.

The objective of this article is to examine the history of the language of development theory in order to elucidate the nature of its terminology. The history of the principal terms of development theory (economic/sustainable/human development, Third World, and North/(Global) South) is examined by way of a quantitative study of the frequency of the usage of these terms during the 20th century based on a dataset of millions of digitised books made available by Google Books. The author argues that the results of the study provide empirical evidence that the language of development theory is a historical-ideological construction which is embedded in the structure of the world economy.

The shifting from a “dilated diasystem”, when Latin becomes more and more complex, to a “disconnected diasystem”, when two distinct linguistic systems appear, have been already described in several models (e.g. Pulgram, 1950 ; Berschin, 1986). Relying on their authors’ extensive experience, these models however leave some problems unadressed. In particular, they consider the language of a specific period as an homogenous entity. They thus mostly ignore the variations of registers existing in the language at a same time, or even in the same text.
In this paper, we propose a methodology to systematically study the evolution of the various registers used in texts written in later Merovingian and in Carolingian ages. This methodology involves computerized statistical analysis and artificial intelligence, but should not be seen as an automated process disconnected from the linguist’s analysis. On the contrary, it has been designed as a means to extend the way of thinking of a particular expert. It enables to partially re-create the point of view of the latter, and to apply it to a large amount of texts, too large to be analyzed otherwise.
The first step consists in the analysis on a selection of texts, aimed at differentiating several registers used inside the texts of our period of interest. Our sample corpus consists in three hagiographical texts and twenty one diplomatic texts. Our three hagiographical texts were written in later Merovingian or in early Carolingian ages (ca. 650-780), then rewritten during the Carolingian Renaissance (from 780 to the death of Charles the Bald). The diplomatic texts are 21 original Frankish royal charters dating from ca. 665 to 868. Most of them are accounting for a judgment.
Following Michel Banniard’s works (Banniard, 2008), we define five language registers, ranked from one to five according to their decreasing proximity to the forms of ancient Latin. Then, we design a table of criteria to characterize them. We eventually analyze every paragraph of our texts, annotating each with the register used.
Using this annotated corpus as training data, we train an artificial intelligence (with the software Weka) to identify the various registers used in a text. The model uses logit regression, applied to n-grams (3 to 6) of characters, words, and n-grams of words (up to 3), and gives satisfying predictions (precision: 74 %). This artificial intelligence is then applied to a large selection of texts written during the same period.
The results show that the registers used tend to slowly, yet significantly, "decrease" during the end of Merovingian times. Then, as expected, the registers used sharply increase. The synchronic variation of registers have a more surprising evolution. While, during Merovingian times, these variations are quite limited, we observe a much greater variability in the registers used during the Carolingian Renaissance. If the "mean" written language is indeed closer to some sermo altus, it also includes passages written in very different registers.

Bibliography (selection):
Michel Banniard, "Du latin des illettrés au roman des lettrés: la question des niveaux de langue en France (viiie-xiie siècle)", in Zwischen Babel und Pfingsten: Sprachdifferenzen und Gesprächsverständigung in der Vormoderne (9.-16. Jh.): Akten der 3. deutsch-französischen Tagung des Arbeitskreises "Gesellschaft und individuelle Kommunikation in der Vormoderne" (GIK) in Verbindung mit dem Historischen Seminar der Univ. Luzern, Höhnscheid (Kassel), 16-19 nov. 2006, Peter von Moos ed., Münster, 2008, pp. 269-286.
W. Berschin, Biographie und Epochenstil im lateinischen Mittelalter, Stuttgart, t. 3: Karolingische Biographie, 750-920, 1991.
Piera Molinelli, "Per una sociolinguistica del latino", in Latin vulgaire – latin tardif: actes du VIIe colloque international sur le latin vulgaire et tardif (Séville, 02-06 septembre 2003), éd. Carmen Arias Abellán, Séville: Universidad de Sevilla, 2006, pp. 463-474.
Giovanni Polara, "Problemi di ortografia e di interpunzione nei testi latini di età carolina, in Grafia e interpunzione del latino nel Medioevo (Roma, sept. 1984), éd. Alfonso Maieru, Rome, 1987, pp. 31-51.
Ernst Pulgram, "Spoken and written Latin", in Language : journal of the Linguistic Society of America, t. 26, 1950, pp. 458-466.

The article explores some implications of recent changes in the English language for the evaluation of student performance. Which changes should the teacher of English as an L2 accept and which she should resist? The article suggests some elements of an answer to these questions.

No matter how strongly linguists believe in descriptivism, as language teachers they cannot avoid being prescriptive when it is their task to mark student work as “right” or “wrong”. However, what they can and indeed must do as university teachers and as scholars of language is to avoid adopting an ex cathedra attitude, taking sole responsibility for valid input. Instead, they should establish a relation of partnership with their students, encouraging them to manage their own learning processes and to find answers to their questions by themselves, outside the classroom.

A potentially useful way of achieving this is to make their students aware of how their language performance is assessed, namely by taking into consideration the variability of contemporary English, subject as it is to ‘normal’ internal pressures to change and adapt (e.g. under grammaticalization or the pull of analogy), the existence of various norms of usage across the different English-speaking parts of the world and the undeniable influence of English as (currently) the primary language of global communication.

The variation of present-day English may be tracked in studies such as Mair’s (2006) and Leech et al.'s (2009), both focusing on recent morphosyntactic changes in this language, or Rohdenburg & Schlüter (2009), given that the tug between British and American norms looms large in teaching practice. In addition, the potential of the Google Ngram Viewer (http://ngrams.googlelabs.com/) may be explored by teacher and students alike, in determining trends in written English.